Google DeepMind's new leader, Koray Kavukcuoglu, used his first media appearance to confirm that Gemini 4 is in its final refinement stage. Speaking to The Information, he said the company intends to release an early post-training output "as soon as possible," aiming for a launch well before the end of 2026. The statement comes after months of speculation and a leadership shakeup that saw Demis Hassabis step down in August.
Google's last flagship model, Gemini 3, arrived in November 2025. Since then, OpenAI's GPT-6 and Anthropic's Mythos series have launched and are widely seen as outperforming it, giving rivals a sizable lead. A planned Gemini 3.5 Pro update, announced at I/O in May, was never rolled out. Kavukcuoglu acknowledged the company "took a little bit of a step back" to focus on faster, less powerful Flash models instead.
Despite the delays, Kavukcuoglu expressed confidence in Google's position, saying, "In my mind, it's a certainty that we are always gonna be at the frontier." The quickened Gemini 4 timeline suggests Google is eager to prove that claim in the near term, though the company has not committed to a specific date. For businesses tracking the AI race, the key question is whether Gemini 4 can match or exceed the capabilities of its newer rivals when it finally ships.